worlabel/ai/app/api/yolo/detection.py
2024-09-25 23:51:19 +09:00

217 lines
7.8 KiB
Python

from fastapi import APIRouter, HTTPException, Request
from schemas.predict_request import PredictRequest
from schemas.train_request import TrainRequest
from schemas.predict_response import PredictResponse, LabelData
from schemas.train_report_data import ReportData
from services.load_model import load_detection_model
from services.create_model import save_model
from utils.dataset_utils import split_data
from utils.file_utils import get_dataset_root_path, process_directories, process_image_and_label, join_path
from utils.slackMessage import send_slack_message
from utils.api_utils import send_data_call_api
import random
router = APIRouter()
@router.post("/predict")
async def detection_predict(request: PredictRequest):
send_slack_message(f"predict 요청: {request}", status="success")
# 모델 로드
model = get_model(request)
# 모델 레이블 카테고리 연결
classes = list(request.label_map) if request.label_map else None
# 이미지 데이터 정리
url_list = list(map(lambda x:x.image_url, request.image_list))
# 추론
results = run_predictions(model, url_list, request, classes)
# 추론 결과 변환
response = [process_prediction_result(result, image, request.label_map) for result, image in zip(results,request.image_list)]
send_slack_message(f"predict 성공{response}", status="success")
return response
# 모델 로드
def get_model(request: PredictRequest):
try:
return load_detection_model(request.project_id, request.m_key)
except Exception as e:
raise HTTPException(status_code=500, detail="load model exception: " + str(e))
# 추론 실행 함수
def run_predictions(model, image, request, classes):
try:
return model.predict(
source=image,
iou=request.iou_threshold,
conf=request.conf_threshold,
classes=classes
)
except Exception as e:
raise HTTPException(status_code=500, detail="model predict exception: " + str(e))
# 추론 결과 처리 함수
def process_prediction_result(result, image, label_map):
try:
label_data = LabelData(
version="0.0.0",
task_type="det",
shapes=[
{
"label": summary['name'],
"color": get_random_color(),
"points": [
[summary['box']['x1'], summary['box']['y1']],
[summary['box']['x2'], summary['box']['y2']]
],
"group_id": label_map[summary['class']] if label_map else summary['class'],
"shape_type": "rectangle",
"flags": {}
}
for summary in result.summary()
],
split="none",
imageHeight=result.orig_img.shape[0],
imageWidth=result.orig_img.shape[1],
imageDepth=result.orig_img.shape[2]
)
except Exception as e:
raise HTTPException(status_code=500, detail="model predict exception: " + str(e))
return PredictResponse(
image_id=image.image_id,
data=label_data.model_dump_json()
)
def get_random_color():
random_number = random.randint(0, 0xFFFFFF)
return f"#{random_number:06X}"
@router.post("/train")
async def detection_train(request: TrainRequest, http_request: Request):
send_slack_message(f"train 요청{request}", status="success")
# Authorization 헤더에서 Bearer 토큰 추출
auth_header = http_request.headers.get("Authorization")
token = auth_header.split(" ")[1] if auth_header and auth_header.startswith("Bearer ") else None
# 레이블 맵
inverted_label_map = {value: key for key, value in request.label_map.items()} if request.label_map else None
# 데이터셋 루트 경로 얻기
dataset_root_path = get_dataset_root_path(request.project_id)
# 모델 로드
model = get_model(request)
# 학습할 모델 카테고리, 카테고리가 추가되는 경우 추가 작업 필요
model_categories = model.names
# 데이터 전처리
preprocess_dataset(dataset_root_path, model_categories, request.data, request.ratio, inverted_label_map)
# 학습
results = run_train(request,token,model,dataset_root_path)
# last 모델 저장
model_key = save_model(project_id=request.project_id, path=join_path(dataset_root_path, "result", "weights", "best.pt"))
response = {"model_key": model_key, "results": results.results_dict}
send_slack_message(f"train 성공{response}", status="success")
return response
def preprocess_dataset(dataset_root_path, model_categories, data, ratio, label_map):
try:
# 디렉토리 생성 및 초기화
process_directories(dataset_root_path, model_categories)
# 학습 데이터 분류
train_data, val_data = split_data(data, ratio)
if not train_data or not val_data:
raise HTTPException(status_code=400, detail="data split exception: data size is too small or \"ratio\" has invalid value")
# 학습 데이터 처리
for data in train_data:
process_image_and_label(data, dataset_root_path, "train", label_map)
# 검증 데이터 처리
for data in val_data:
process_image_and_label(data, dataset_root_path, "val", label_map)
except HTTPException as e:
raise e # HTTP 예외를 다시 발생
except Exception as e:
raise HTTPException(status_code=500, detail="preprocess dataset exception: " + str(e))
def run_train(request, token, model, dataset_root_path):
try:
# 데이터 전송 콜백함수
def send_data(trainer):
try:
# 첫번째 epoch는 스킵
if trainer.epoch == 0:
return
# 남은 시간 계산(초)
left_epochs = trainer.epochs - trainer.epoch
left_seconds = left_epochs * trainer.epoch_time
# 로스 box_loss, cls_loss, dfl_loss
loss = trainer.label_loss_items(loss_items=trainer.loss_items)
data = ReportData(
epoch=trainer.epoch, # 현재 에포크
total_epochs=trainer.epochs, # 전체 에포크
box_loss=loss["train/box_loss"], # box loss
cls_loss=loss["train/cls_loss"], # cls loss
dfl_loss=loss["train/dfl_loss"], # dfl loss
fitness=trainer.fitness, # 적합도
epoch_time=trainer.epoch_time, # 지난 에포크 걸린 시간 (에포크 시작 기준으로 결정)
left_seconds=left_seconds # 남은 시간(초)
)
# 데이터 전송
send_data_call_api(request.project_id, request.m_id, data, token)
except Exception as e:
raise HTTPException(status_code=500, detail=f"send_data exception: {e}")
# 콜백 등록
model.add_callback("on_train_epoch_start", send_data)
# 학습 실행
try:
results = model.train(
data=join_path(dataset_root_path, "dataset.yaml"),
name=join_path(dataset_root_path, "result"),
epochs=request.epochs,
batch=request.batch,
lr0=request.lr0,
lrf=request.lrf,
optimizer=request.optimizer
)
except Exception as e:
raise HTTPException(status_code=500, detail=f"model train exception: {e}")
# 마지막 에포크 전송
model.trainer.epoch += 1
send_data(model.trainer)
return results
except HTTPException as e:
raise e # HTTP 예외를 다시 발생
except Exception as e:
raise HTTPException(status_code=500, detail=f"run_train exception: {e}")